该博客讨论了遇到的TensorFlow内部错误，具体表现为BlasGEMM启动失败。作者提供了通过设置GPU内存增长来解决此问题的方案，即通过`tf.config.list_physical_devices('GPU')`和`tf.config.experimental.set_memory_growth`来动态管理GPU内存。
错误信息如下：
tensorflow.python.framework.errors_impl.InternalError: 2 root error(s) found.
(0) Internal: Blas GEMM launch failed : a.shape=(10, 10), b.shape=(10, 10), m=10, n=10, k=10
[[{{node sequential/simple_rnn/while/body/_1/MatMul_1}}]]
[[Reshape_6/_32]]
(1) Internal: Blas GEMM launch failed : a.shape=(10, 10), b.shape=(10, 10), m=10, n=10, k=10
[[{{node sequential/simple_rnn/while/body/_1/MatMul_1}}]]
解决方案：
physical_devices = tf.config.list_physical_devices('GPU')
tf.config.experimental.set_memory_growth(physical_devices[0], True)
扫一扫
1万+
到【灌水乐园】发言
